World Applied Sciences Journal 26 (1): 75-82, 2013 ISSN 1818-4952 © IDOSI Publications, 2013 DOI: 10.5829/idosi.wasj.2013.26.01.1323

Water Quality of River, Lake and Drinking Water Supply in State by Means of Multivariate Analysis

Suhair S. Salih, Abbas F.M. Alkarkhi, Japareng Bin Lalung and Norli Ismail

School of Industrial Technology, Universiti Sains , 11800 Penang, Malaysia

Submitted: Sep 8, 2013; Accepted: Nov 7, 2013; Published: Nov 10, 2013 Abstract: Statistical techniques such as multivariate analysis of variance (MANOVA) and discriminate analysis (DA) were used to analyze the data obtained from three different locations (rivers, lake, drinking water supply). Seven parameters were measured pH, temperature, TSS, COD, BOD, turbidity and E. coli to investigate the pollution status. MANOVA showed a strong significant difference. While discriminate analysis (DA) explained the differences between different locations with the use of two functions. The first function showed 98.4% total variation, mainly due to E. coli, turbidity, BOD, temperature and COD. Even as the second function recorded 1.6% total variation, mainly due to E. coli, COD, temperature and pH. DA was also used to determine the relative contribution for all parameters in differentiating between river, lake and tap water. The results also showed strong correlations between COD and suspended solids. BOD with temperature, COD and pH. Turbidity with pH, temperature, COD and BOD. E. coli with COD, BOD, pH, temperature and suspended solids and a strong association of temperature with pH. These derived relationships can be used to produce reasonable planning estimates for the frequency of high E. coli levels (>200/100 mL) in rivers and lakes.

Key words: MANOVA Discriminate analysis E. coli BOD COD

INTRODUCTION necessary to forestall and control the rivers pollution by acquiring reliable data water quality for effective There has been a universal interest about faecal management. contamination of natural water bodies and its potential Penang is a state located in the northwest of impact on human health. Several studies on Best Malaysia. The state has six dam structures managed by Management Practices (BMPs) have been conducted to PBA (Perbadanan Bekalan Air). Between these reservoirs, reduce untreated human and domestic animal waste Mengkuang Reservoir (M.R) has the highest capacity. discharge, however, our limited understanding of bacteria The capacity of Mengkuang Reservoir is almost nine dynamics within a watershed remains imperfect [1, 2]. times more than Reservoir. The present water Rivers have become a source of tourist attraction and supplies are sufficient for Penang state, though increasing hospitality, which has prompted the construction of population demands proactive steps by the government hotels and resorts around the area. These drastic to effectively manage and conserve the existing water developments have caused deforestation, resulting in supply. The surrounding area of Mengkuang Reservoir gradual erosion of the surrounding bare soil during consists of natural tropical forest and recreational spots rainfall. The weathered soil runs into the rivers and forms comprising landscaped gardens and sitting benches. a murky layer that blocks out sunlight from reaching the M.R is a part of the Mengkuang pumped storage scheme aquatic life thereby causing the loss of marine life. that supplies more than 81 million gallons of untreated Since, rivers constitute the main inland water resources and drinking water per day to , Butterworth for domestic, industrial and irrigation purposes, it is and other parts of the state. Recently, there has been an

Corresponding Author: Norli Ismail, School of Industrial Technology, Universiti Sains Malaysia, 11800 Penang, Malaysia. 75 World Appl. Sci. J., 26 (1): 75-82, 2013 increase in economic development in Malaysia as a result E) and flows through the state. The total area of the of more land use, urbanization, industrialization and the Sungai basin is about 50.97 km2 and its length expansion of mechanized agriculture. These economic is approximately 3.1 km. Over half of the basin is covered growth factors have adversely affected the quantity and with forest with about one-third of the river basin area quality of water supply while simultaneously increasing being used as arable land. The water quality of Sungai the demand for limited freshwater resources. Therefore, Pinang is categorized as Class V which is considered appropriate management and monitoring of drinking water "very polluted", in accordance with proposed Interim supply is critical. National Water Quality Standards (INWQS) for Monitoring of physio-chemical parameters is a Malaysia. Figure 1 shows the location of routine water quality assessment for drinking water River. supply in Malaysia. While chemical analysis has certain Specifically, the two lakes in Universiti Sains limitations such as time, cost and technical restraints, Malaysia (USM) main campus; Lake Tasik Harapan biological studies are capable of providing continuous (0.80 hectares) and Lake Aman approximately (0.58 acres), temporal and spatial information on water ecosystem were chosen for this study. These are one of the without aforementioned limitations. In Malaysia, man-made engineering landscapes that have garnered biological monitoring in water supplies normally involves much attention from the community in the campus. total coli form count for detection of focal pollution. [3] The lakes were constructed in 1990 as retention ponds in The application of multivariate methods for analyzing order to reduce the effect of flash flood after a heavy environmental data has increased enormously in recent downpour. To maintain the beauty of these lakes, a times [4, 5, 6]. Multivariate methods involve the 20 meter buffer area around the lake was provided [12]. measurement of several dependant variables for each However, from recent site observations, it was found that sampling unit. Multivariate analysis of variance the lakes are severely polluted and the water is greenish (MANOVA) has been used to test the significant in colour [13]. differences, while discriminate function (DF) has been used to identify the relative contribution of all variables to Water Sampling and Analysis: The technique used for the separation of the groups [7, 8]. the collection of water samples is referred to as grab Water pollution is going to be a serious problem in sampling. Water samples were taken at each tributary Malaysia and post a negative impacts on the using a bucket attached to a rope. Samples for laboratory sustainability of water resources plants and living analyzing were collected in HDPE bottles that were organisms, people's health and the country's economy [9]. pre-soaked in HCI for 24 hours and copiously rinsed with Previous studies relevant to the water quality of Sungai deionizer water. The sample containers were filled slowly Pinang Basin explained that there are several major water to the brim to avoid air bubbles. Samples collected were pollution sources attributed to anthropogenic activities then stored at 4°C. The samples were collected out during such as domestic sewage, agricultural and industrial the dry weather. wastes [10, 11]. The objective of this research project is to determine the differences in pollution status between Parameters Selected for Analysis: Samples obtained river, lake and drinking water supply, based on several from different locations were analyzed for seven parameters which include; pH, temperature, TSS, COD, parameters: pH, temperature °C, Biochemical Oxygen BOD, turbidity and E. coli, in order to identify differences Demand (BOD) (mg/l), Chemical Oxygen Demand (COD) in pollution between these locations. (mg/l), Total Suspended Solids (TSS) (mg/l), turbidity (NTU) and E. coli (CFU/100 ml). The methods of analysis MATERIALS AND METHODS are in line with the requirements of Standard Methods [14]. Temperature and pH were measured in site by using Background: The study site is located within Sungai thermometer and pH meter( ph107 Digital Pocket PH Meter Pinang River basin in the state of Penang, on the North tester) respectively. While the other parameters such as West coast of Peninsular Malaysia, [10]. It also comprises Biochemical Oxygen Demand, Chemical Oxygen Demand, the Sungai Pinang river, which is situated at the Total Suspended Solids, turbidity and E. coli were carried North-Eastern part of the Penang Island (5° 24' N l00° l9' out in the laboratory.

76 World Appl. Sci. J., 26 (1): 75-82, 2013

Fig. 1: Map of Sungai Pinang River

RESULTS AND DISCUSSION The maximum values of turbidity showing the highest values for lake and river samples with values of Descriptive Statistics: Descriptive statistics for the 83.00 and 55.00 respectively. Whereas turbidity of selected parameters, which include: minimum, maximum, drinking water supply exhibited the lowest value with 0.13. mean and standard deviation for different locations river, As the minimum turbidity values of lake, river and lake and drinking water supply are shown in Table 1. drinking water supply were 77.00, 19.00 and 0.04 The mean value of E. coli was arranged for three locations respectively. High turbidity values could be due to in the following order: river > lake > drinking water supply. possible pollution by soil erosion from the construction The E. coli sample from river displayed the highest level and earthworks activities within this area [10]. The greater with 1162.5000, whilst E. coli value for the lake showed the amount of turbidity in the lake, the murkier it appears lower value compared to river 211.25, although the E. coli and the higher the measured total suspended solids sample from drinking water supply exhibited the lowest (TSS). The major source of turbidity in the open water value with 2.2500. On the other hand, the minimum values zone of most lakes is typically phytoplankton [17, 18]. of the E. coli samples from river, lake and drinking water The main cause of pollution in Lake Tasik Harapan supply were, 500.00, 170.00 and 0.00 respectively. The and Lake Aman is the poor quality of storm water which difference in E. coli values could be due to the effect of flows into both lakes through Sungai Gambir. The storm industrial discharge such as food processing plants, water manages to flow through the control gate at the abattoirs, residential areas and wet market, which inlet when the level of water increased. The lake is also discharging their wastes into the river [15], additionally polluted with the effluent discharged from the Indah degradation within the lake drainage basin caused by Water Konsortium (IWK) wastewater treatment plant. excess sediment and non- point sources inputs associated This situation has contributed to the loss of ecological with rapid development, increased land use and pollution functions and destruction of ecosystem in the lake. from agrochemical and sewage effluent affect the E. coli Figure 2 shows the location of Lake Hope in USM main values [16]. campus.

77 World Appl. Sci. J., 26 (1): 75-82, 2013

of water. This an indicator of untreated effluent discharged into these water bodies, such as domestic sewage, agricultural and industrial wastes. The correlation matrix of the concentration of selected parameters measured from different locations, namely river, lake and drinking water supply were examined. The correlation coefficients for selected parameters are given in Table 2. For water sampled from river, a strong negative relationship was exhibited between COD and E. coli with TSS and between turbidity with pH and BOD and positive relationship between BOD and pH and between E. coli and COD. Different patterns of relationship were shown Fig. 2: Map of Tasik Harapan Lake in USM between the selected parameters for water sampled from a lake since a strong positive relationship was exhibited Table 1: Descriptive statistics including Minimum, Maximum, Mean and between E. coli with temperature, pH and BOD and Std. Deviation for the different locations river, lake and drinking water supply between COD, BOD with pH and temperature and positive River relationship was displayed between temperature and pH, Descriptive Statistics whereas there is negatively relationship between turbidity N Minimum Maximum Mean Std. Deviation with temperature, COD, BOD and pH and positively pH 4 6.50 7.50 7.1250 .47871 relationship between BOD and COD for drinking water Temp 4 29.00 32.00 30.0000 1.41421 supply. TSS 4 16.00 37.00 27.0000 10.55146 Although bacteria-TSS relationships have been COD 4 200.00 1100.00 650.0000 465.47467 previously examined separately by different research BOD 4 7.00 9.50 8.5000 1.08012 Turb 4 19.00 55.00 31.0000 16.75311 teams [19,20], sufficiently strong relationships between Ecoli1 4 500.00 2000.00 1162.5000 711.07313 bacteria, turbidity and TSS may be discovered and used Lake to develop planning level timales of indicator bacteria Descriptive Statistics (e.g. fecal coli form) levels on a continuous basis. This N Minimum Maximum Mean Std. Deviation could help to decrease the costs and time-constraints of pH 4 6.20 6.70 6.4750 .20616 fecal coli form sampling. Oxygen-related parameters, COD Temp 4 26.00 27.00 26.5000 .40825 and BOD, showed high positive correlation with E. coli. TSS 4 83.00 140.00 120.2500 25.38208 COD is a very important indicator for E. coli growth as it COD 4 90.00 100.00 96.1000 4.27863 BOD 4 10.00 12.00 11.0750 .83016 measures the capacity of water to consume oxygen Turb 4 77.00 83.00 79.7500 2.50000 during the decomposition of organic matter and the Ecoli1 4 170.00 250.00 211.2500 34.24787 oxidation of inorganic chemicals such as ammonia and Drinking Water Supply nitrate. BOD also showed a strong linear correlation with Descriptive Statistics E. coli, therefore indicating that the water contained N Minimum Maximum Mean Std. Deviation relatively high concentration of sewage contamination. pH 4 6.50 7.00 6.6500 .23805 Also there is positively correlation between pH and Temp 4 25.00 26.00 25.4750 .55000 temperature with COD and BOD, when the degree of pH TSS 4 .00 .00 .0008 .00056 COD 4 .00 3.00 1.8750 1.31498 is very low or high, for both cases the growth of bacteria BOD 4 .00 1.00 .7500 .50000 was decrease. Nevertheless, in low temperature, the cell Turb 4 .04 .13 .0856 .04058 metabolism starts to slow while high temperature leads to E. coli 4 .00 5.00 2.2500 2.21736 stop the growth of bacteria [21]. That means the temperature and pH are effected on the COD and BOD. The resulting values of the several parameters The suspended particles in water which causes the analyzed for the river and lake samples are higher than the turbidity, adsorb the sunlight and lead to increase the acceptable limit established by Interim National Water temperature of water [22], this explains the relationship Quality Standard (INWQS) for Malaysia while the between turbidity and the temperature of water. parameter values for the drinking water supply are within Also when the pH of water was low, the removal ratio the allowable range. Therefore, The Sungai Pinang River of turbidity will increase [23].This indicates that and Lake Tasik Harapan are considered polluted sources relationship between turbidity and pH was negatively.

78 World Appl. Sci. J., 26 (1): 75-82, 2013

Table 2: Correlation matrix for selective parameter in Sungai Pinang, Tasik Harapan and tap water. River pH Temp. TSS COD BOD Turbidity E. coli pH 1 Temp. .739 1 TSS -.924 -.894 1 COD .935 .861 -.998** 1 BOD .967* .764 -.863 .862 1 Turbidity -.977 -.661 .832 -.842 -.986* 1 E. coli .900 .945 -.991** .979* .868 -.816 1 Lake pH Temp. TSS COD BOD Turbidity E. coli pH 1 Temp. .990** 1 TSS -.852 -.917 1 COD .986* .954* -.758 1 BOD .988* .984* -.865 .975* 1 Turbidity -.986* -.980* .858 -.975* -1.000** 1 E. coli .950* .954* -.867 .934 .986* -.988* 1 Trap pH Temp. TSS COD BOD Turbidity E. coli pH 1 Temp. .777 1 TSS -.804 -.819 1 COD .612 .570 -.933 1 BOD .420 .576 -.875 .951* 1 Turbidity -.805 -.974* .926 -.740 -.730 1 E. coli .789 .444 .833 -.872 .676 -.604 1 * Correlation is significant at the 0.05 level (2-tailed ). **Correlation is significant at the 0.01 level (2-tailed ).

Table 3: Multivariate test (MANOVA) for all locations Effect Value F Sig. Location Pillai's Trace 1.939 18.174 0.000 Wilks' Lambda 0.000 53.312b 0.000 Hotelling's Trace 956.651 136.664 0.000 Roy's Largest Root 940.959 537.691c 0.000

Multivariate Analysis: Multivariate analysis of Discriminate analysis consisted of seven parameters. variance (MANOVA) was used to determine the Two discriminate functions were responsible for difference in the water samples collected from three differentiating the three locations as shown in Eqs. 1 locations based on seven parameters. The results of and 2. Wilk’s Lamda test showed that DF is statistically MANOVA showed differences in the concentrations of significant as indicated by p<0.0001 in Table 4. selected parameter in water sampled from the three Furthermore, 100% of the total variation between the three locations (Table 3), which suggests that the source of locations was explained by two discriminate functions. pollution is different for each location. For instance, The relative contribution of each parameter is given in domestic sewage, agricultural and industrial wastes, soil Eqs. 1 and 2. erosion from construction site and urban runoff are identified as the main sources of pollution in river [24], Z = 0.65 pH - 2.28 Temp. + 0.96 TSS - 1.01 COD + 2.75 BOD while pollution in lake include eutrophication, inorganic + 3.09 Trub. + 3.53 E. coli pollution sedimentation and siltation, over-exploitation, (1) loss of biodiversity, habitat change, aquatic plant infestation [16]. Z = 2.40 pH - 3.61 Temp. + 0.02 TSS -5.75 COD -0.41 BOD Spatial variations in selected parameter + 1.06 Trub. + 7.69 E. coli concentrations were further evaluated using DA. (2)

79 World Appl. Sci. J., 26 (1): 75-82, 2013

Table 4: Wilks’ Lambda for testing discriminate function validity Wilks' Lambda Test of Function(s) Wilks' Lambda Chi-square df Sig. 1 through 2 0.000 57.978 14 <0.0001 2 0.060 16.890 6 <0.010

Table 5: Eigen-value for each discriminate function Function Eigenvalue % of Variance Cumulative % Canonical Correlation 1 940.959a 98.4 98.4 0.999 2 15.693a 1.6 100.0 0.970 a. First 2 canonical discriminant functions were used in the analysis.

D1

40

30

20

10

0 D1 RRRRLLLLTTTT

-10

-20

-30

-40 Fig. 3: Scores for the first standardized discriminate function D2

4

2

0 RRR R L L L L TTT T

-2 D2

-4

-6

-8 Fig. 4: Scores for the second standardized discriminate function

80 World Appl. Sci. J., 26 (1): 75-82, 2013

The first discriminate function explains 98.4 % total REFERENCES variation between different locations whilst the second discriminate function explains 1.6 % total variation. Eq. 1 1. Geldreich, E., 1989. ‘Pathogens in freshwaters. showed that E. coli contributed highly in discriminating Globa Freshwater Quality, a First Assessment, the three locations and account for most of the pp: 58-78. expected variations in Eq. 1, followed by turbidity 2. Embrey, S.S., 2001. ‘Microbiological Quality of which provided the second highest contribution, Puget Sound Basin Streams and Identification of while other parameters contributed less in discriminating Contaminant Sources1’. JAWRA Journal of the the locations. Eq. 2 showed that E. coli contributed highly American Water Resources Association, 37: 407-421. in discriminating the three locations and accounted for 3. Makhlough, A., 2008. ‘Water Quality Characteristics most of the expected variations in Eq. 2, with COD Of Mengkuang Reservoir Based On Phytoplankton showing the second highest support while other Community Structure And Physico-Chemical parameters contributed less in differentiating the Analysis’ [TD370. M235 2008 f rb], Universiti Sains locations. Malaysia. The relationship between the scores of discriminate 4. Tuncer, G.T., S.G. Tunce, G. Tuncel and T.I. Balkas, function and water sampled from various locations are 1993.‘Metal pollution in the Golden Horn, Turkey- presented in Figs. 3 and 4, which correspond to the contribution of Natural and Anthropogenic values of discriminate function for the different samples. Components Since 1913’. The samples R represents water samples collected from 5. Einax, J. and U. Soldt, 1999. ‘Geostatistical and the river, L represents water sample of lake and T water multivariate statistical methods for the assessment of samples of drinking water supply. Fig. 3 reveals that water polluted soils-merits and limitations’. Chemo Metrics sampled from drinking water supply contributed and Intelligent Laboratory Systems, 46: 79-91. negatively while water sampled from river and lake 6. Abd-Wahab, S.A., C.S. Bakheit and S.M. Al-Alawi, contributed positively to the first discriminate function. 2005. ‘Principal component and multiple regression Fig. 4 indicates that water sampled from river contributed analysis inmodelling of ground level ozone and negatively while water sampled from lake and drinking factors affecting its concentrations’ Environmental water supply contributed positively to the second Modelling and Software, 20: 1263-1271. discriminate function. These differences are related to the 7. Richard, A.J. and W.W. Dean, 2002. ‘Applied reasons many toxic chemicals and acids entered the lakes multivariate statistical analysis’. Englewood Cliffs, and reservoirs from the atmosphere and pollution often NJ: Prentice-Hall. occurs on lakes rather than streams or rivers. TSS and 8. Alvin, C.R., 2002. Methods of multivariate analysis. turbidity concentrations are the main pollutants for lakes Wiley. which affect the water clarity. 9. Jacky, B. Ling, 2010. ‘Water quality study and its relationship with high tide and low tide at Kuantan CONCLUSION river’. Universiti Malaysia Pahang. 10. DID, Y., 2000. ‘ Quality Monitoring Report for Sungai Based on the above results and discussion it can be Pinang River Basin, Drainage and Irrigation seen that multivariate statistical technique has been Department’ Penang. proved to be an effective tool for discriminating different 11. Koh, H.L., 2004. ‘PemodelanAlam Sekilar dan locations and identifying the relative contribution of each Ekosistem’ Penerbit Universiti Sains Malaysia. parameter as well. Most of the parameters tested on the 12. Dzulkifli Abdul Razak.,2001.‘The Concept Plan Pelan water samples from Sungai Pinang River fall within Class Konsep’. http://www.hbp.usm.my/usm/TheProject/ V, while E. coli levels (>200 /100mL) and other parameters ConceptPlan/siteanalysis/SiteAnalysis2.htm. for Lake Tasik Harapan fall within Class IV, but drinking 13. Zorkeflee Abu Hasan, Nuramidah Hamidon, Norazazi water supply is within the allowable range of established Zakaria, Aminuddin Ab. Ghani and Leow Cheng by Interim National Water Quality Standard (INWQS) Siang, Universiti Sains Malaysia, Engineering for Malaysia. Therefore, Sungai Pinang River can be Campus, Seri Ampangan, 14300 , classified as very polluted. Penang, Malaysia. Email: redac02@eng.

81 World Appl. Sci. J., 26 (1): 75-82, 2013

14. APHA, 1992. Standard Methods for the Examination 20. Irvine, K.N., G.W. Pettibone and I.G. Droppo, 1995. of Water and Wastewater (1&h ed.), American Public ‘Indicator Bacteria-Sediment Relationships: Health Association, Washington, DC. Implications for Water Quality Modelling and 15. Ruslan, H., 1998. ‘Holistic Approach to Urban Storm- Monitoring’. In Modern Methods for Modeling the ll/Water Pollution Management, Buletin Ingenious, Management of Storm water Impacts, ed. W. James, Kuala Lumpur. ch. 14. Ann Arbor: Ann Arbor Science. 16. Zati Sharip and Salmah Zakaria, 2007. National 21. Yuxin Zhu, 2007. The Effects of pH and Temperature Overview. ‘The status of lakes eutrophication in on the Growth of Escherichia coli DH5a California Malaysia’, pp: 24-34. State Science Fair. 17. Moore, M.L., 1989. NALMS management guide for 22. Paaijmans, K.P., W. Takken, A.K. Githeko and lakes and reservoirs. North American Lake A.F. Jacobs, 2008. The effect of water turbidity on the Management Society, P.O. Box 5443, Madison, WI, near-surface water temperature of larval habitats of 53705-5443, USA. the malaria mosquito Anopheles Gambiae. 18. Michaud, J.P., 1991. A citizen's guide to 23. Ahamed Fadel Ashery, Kamal Radwan and I. understanding and monitoring lakes and streams. Mohamed Gar Al-Alm Rashed, 2010. Alexandria, Publ. #94-149. Washington State Department of Egypt The effect of pH control on turbidity and NOM Ecology, Publications Office, Olympia, WA, USA Removal in Conventional Water Treatment. 360-407-7472. 24. Tan, K.W., 2005. ‘Point Source (Factory and Sewage 19. Sun, H., P. Cornish and T.M. Daniell, 2001. Treatment Plant)’ Water Quality Characterization ‘Turbidity-based erosion estimation in a Study of Sungai Kecil, Sungai Air Terjun, Sungai catchment in South Australia. Journal of Hydrology’, Jelutong and Sungai Pinang, Universiti Sains 253: 227-238. Malaysia, Minden.

82